University Ranking Analysis¶
Analysis is done for the rankings given by
- Times Higher Education (THE)
- Center for World University Rankings (CWUR)
Analysis has been done in order to find the factors that drive their universities' ranking and also to compare their output.
Major insights from the analysis:
- CWUR ranking is highly based on quality of education, quality of faculty, citations and influence. Other variables used in CWUR ranking like patents, publications and alumni employment also affect world ranking.
- THE ranking is mainly determined by teaching, research and citations.THE ranking also lists female male ration, student staff ratio and number of students but these factors do not have much effect on ranking.
- Both CWUR and THE rankings have many institutions from USA, UK, Japan, France, Germany and China in their list.
- If we consider just the top 10 institutes then most of the institutes are form the USA followed by UK for most of the factors. For few factors institutes from Japan, Fance and Netheralnds are also in top 10 institutes. Massachusetts Institute of Technology, Harvard University, Stanford University and Princeton University in USA; University of Oxford and University of Cambridge in UK have been consitently in the top.
In [99]:
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
%matplotlib inline
from wordcloud import WordCloud, STOPWORDS, ImageColorGenerator
import plotly.plotly as py
import plotly.graph_objs as go
from plotly.offline import download_plotlyjs, init_notebook_mode, plot, iplot
In [2]:
times_data = pd.read_csv('C:/Users/Vijaya/Data Science/Viz/University_data/timesData.csv')
shanghai_data = pd.read_csv('C:/Users/Vijaya/Data Science/Viz/University_data/shanghaiData.csv')
cwur_data = pd.read_csv('C:/Users/Vijaya/Data Science/Viz/University_data/cwurData.csv')
CWUR Ranking Analysis¶
In [3]:
cwur_data.head()
Out[3]:
In [4]:
cwur_data.info()
In [5]:
cwur_data.describe()
Out[5]:
In [6]:
cwur_data.shape
Out[6]:
In [7]:
cwur_data.dtypes
Out[7]:
In [8]:
cwur_data.year.unique()
Out[8]:
Ranking of universities per year¶
CWUR started ranking the universities in 2012. Years 2012 and 2013 had top 100 universities rated. From year 2014 1000 universities are being rated
In [9]:
sns.countplot(x='year',data=cwur_data)
Out[9]:
Countries having top universities¶
All 4 years (2012-2015) it is the USA which has the maximum number of top universities in the world. It is followed by United Kingdom in years 2012-13 and by China in years 2014-15. So, among top 100 United Kingdom has more universities than China but among top 1000 China has more universities. China is a huge country as compared to United Kingdom both in terms of area and poulation. Other countries having top rated universities are Japan, France, Germany and Israel.
In [10]:
cntry_inst_2012 = cwur_data[cwur_data['year']==2012]['country'].value_counts().head(15)
cntry_inst_2012 = pd.DataFrame(cntry_inst_2012)
cntry_inst_2012.rename(columns={'country': 'count'}, inplace=True)
cntry_inst_2012['country'] = cntry_inst_2012.index
cntry_inst_2013 = cwur_data[cwur_data['year']==2013]['country'].value_counts().head(15)
cntry_inst_2013 = pd.DataFrame(cntry_inst_2013)
cntry_inst_2013.rename(columns={'country': 'count'}, inplace=True)
cntry_inst_2013['country'] = cntry_inst_2013.index
cntry_inst_2014 = cwur_data[cwur_data['year']==2014]['country'].value_counts().head(15)
cntry_inst_2014 = pd.DataFrame(cntry_inst_2014)
cntry_inst_2014.rename(columns={'country': 'count'}, inplace=True)
cntry_inst_2014['country'] = cntry_inst_2014.index
cntry_inst_2015 = cwur_data[cwur_data['year']==2015]['country'].value_counts().head(15)
cntry_inst_2015 = pd.DataFrame(cntry_inst_2015)
cntry_inst_2015.rename(columns={'country': 'count'}, inplace=True)
cntry_inst_2015['country'] = cntry_inst_2015.index
In [11]:
# Country wise institutions ranked count
fig = plt.figure(figsize=(14,10))
ax1 = fig.add_subplot(221)
ax2 = fig.add_subplot(222)
ax3 = fig.add_subplot(223)
ax4 = fig.add_subplot(224)
plt.subplots_adjust(hspace=0.7)
g = sns.barplot(x='country',y='count',data=cntry_inst_2012, ax=ax1)
g.axes.set_title('2012', fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
h = sns.barplot(x='country',y='count',data=cntry_inst_2013, ax=ax2)
h.axes.set_title('2013', fontsize=20,color="r",alpha=0.5)
for item in h.get_xticklabels():
item.set_rotation(90)
i = sns.barplot(x='country',y='count',data=cntry_inst_2014, ax=ax3)
i.axes.set_title('2014', fontsize=20,color="r",alpha=0.5)
for item in i.get_xticklabels():
item.set_rotation(90)
j = sns.barplot(x='country',y='count',data=cntry_inst_2015, ax=ax4)
j.axes.set_title('2015', fontsize=20,color="r",alpha=0.5)
for item in j.get_xticklabels():
item.set_rotation(90)
Countries having top rated universities in patents¶
These graphs are produced by averaging the patents rating i.e. sum of rating for all the institutions of a country per year divided by number of institutions for that country that got ranked. Countries like Singapore, Israel, Malaysia and Denmark are in top while rating for patents as there are only couple of institutions for these countries and they are ranked in top 100.
In [12]:
# Country wise average patents ranking per institute
ctry_patents_avg = cwur_data.groupby(['year','country'])['patents'].mean()
ctry_patents_avg = pd.DataFrame(ctry_patents_avg)
ctry_patents_avg = ctry_patents_avg.reset_index()
ctry_patents_avg = ctry_patents_avg.sort_values(['year','patents'])
ctry_patents_avg = ctry_patents_avg.rename(columns = {'patents':'avg_patents'})
fig = plt.figure(figsize=(17,10))
ax = []
ax.append(fig.add_subplot(221))
ax.append(fig.add_subplot(222))
ax.append(fig.add_subplot(223))
ax.append(fig.add_subplot(224))
plt.subplots_adjust(hspace=0.7)
years = cwur_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_patents',data=ctry_patents_avg[ctry_patents_avg['year'] == y].head(20), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries having top education quality¶
Countries like Russia, Switzerland, Lebanon, Bulgaria and Belgium have instititions that are rated highly in quality of education. These have only few institutions and they are ranking high due to which they are rated highly. USA, United Kingdom, France and Germany are other countries that have many instututions and are rated high for quality of education.
In [13]:
# Country wise average quality of education per institute
ctry_qoe_avg = cwur_data.groupby(['year','country'])['quality_of_education'].mean()
ctry_qoe_avg = pd.DataFrame(ctry_qoe_avg)
ctry_qoe_avg = ctry_qoe_avg.reset_index()
ctry_qoe_avg = ctry_qoe_avg.sort_values(['year','quality_of_education'])
ctry_qoe_avg = ctry_qoe_avg.rename(columns = {'quality_of_education':'avg_quality_of_education'})
fig = plt.figure(figsize=(17,10))
ax = []
ax.append(fig.add_subplot(221))
ax.append(fig.add_subplot(222))
ax.append(fig.add_subplot(223))
ax.append(fig.add_subplot(224))
plt.subplots_adjust(hspace=0.7)
years = cwur_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_quality_of_education',data=ctry_qoe_avg[ctry_qoe_avg['year'] == y].head(15), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries having top alumni employment¶
Countries like South Africa, Colombia, Lebanon and Russia have instititions that are rated highly in alumni employment. These have only few institutions and they are ranking high due to which they are rated highly. USA, United Kingdom and Canada are other countries that have many instututions and are rated high for alumni employment.
In [14]:
# Country wise average alumni employment per institute
ctry_alEmp_avg = cwur_data.groupby(['year','country'])['alumni_employment'].mean()
ctry_alEmp_avg = pd.DataFrame(ctry_alEmp_avg)
ctry_alEmp_avg = ctry_alEmp_avg.reset_index()
ctry_alEmp_avg = ctry_alEmp_avg.sort_values(['year','alumni_employment'])
ctry_alEmp_avg = ctry_alEmp_avg.rename(columns = {'alumni_employment':'avg_alumni_employment'})
fig = plt.figure(figsize=(17,10))
ax = []
ax.append(fig.add_subplot(221))
ax.append(fig.add_subplot(222))
ax.append(fig.add_subplot(223))
ax.append(fig.add_subplot(224))
plt.subplots_adjust(hspace=0.7)
years = cwur_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_alumni_employment',data=ctry_alEmp_avg[ctry_alEmp_avg['year'] == y].head(15), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries having top faculty quality¶
Countries like Israel, Switzerland, Sweden Denmark and Norway have instititions that are rated highly in quality of faculty. These have only few institutions and they are ranking high due to which they are rated highly. USA and United Kingdom are other countries that have many instututions and are rated high for faculty quality.
In [15]:
# Country wise average quality_of_faculty per institute
ctry_qof_avg = cwur_data.groupby(['year','country'])['quality_of_faculty'].mean()
ctry_qof_avg = pd.DataFrame(ctry_qof_avg)
ctry_qof_avg = ctry_qof_avg.reset_index()
ctry_qof_avg = ctry_qof_avg.sort_values(['year','quality_of_faculty'])
ctry_qof_avg = ctry_qof_avg.rename(columns = {'quality_of_faculty':'avg_quality_of_faculty'})
fig = plt.figure(figsize=(17,10))
ax = []
ax.append(fig.add_subplot(221))
ax.append(fig.add_subplot(222))
ax.append(fig.add_subplot(223))
ax.append(fig.add_subplot(224))
plt.subplots_adjust(hspace=0.7)
years = cwur_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_quality_of_faculty',data=ctry_qof_avg[ctry_qof_avg['year'] == y].head(15), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries rated top in publications¶
Countries like Singapore, Netherlands, Sweden, Denmark and Hong Kong have instititions that are rated highly in publications. These have only few institutions and they are ranking high due to which they are rated highly. USA and Canada are other countries that have many instututions and are rated high for publications.
In [16]:
# Country wise average publications per institute
ctry_publications_avg = cwur_data.groupby(['year','country'])['publications'].mean()
ctry_publications_avg = pd.DataFrame(ctry_publications_avg)
ctry_publications_avg = ctry_publications_avg.reset_index()
ctry_publications_avg = ctry_publications_avg.sort_values(['year','publications'])
ctry_publications_avg = ctry_publications_avg.rename(columns = {'publications':'avg_publications'})
fig = plt.figure(figsize=(17,10))
ax = []
ax.append(fig.add_subplot(221))
ax.append(fig.add_subplot(222))
ax.append(fig.add_subplot(223))
ax.append(fig.add_subplot(224))
plt.subplots_adjust(hspace=0.7)
years = cwur_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_publications',data=ctry_publications_avg[ctry_publications_avg['year'] == y].head(15), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries rated top in citations¶
Countries like Singapore, Netherlands, Sweden, Denmark and Switzerlans have instititions that are rated highly in citations. These have only few institutions and they are ranking high due to which they are rated highly. USA, United Kingdom and Canada are other countries that have many instututions and are rated high for citations.
In [17]:
# Country wise average citations per institute
ctry_citations_avg = cwur_data.groupby(['year','country'])['citations'].mean()
ctry_citations_avg = pd.DataFrame(ctry_citations_avg)
ctry_citations_avg = ctry_citations_avg.reset_index()
ctry_citations_avg = ctry_citations_avg.sort_values(['year','citations'])
ctry_citations_avg = ctry_citations_avg.rename(columns = {'citations':'avg_citations'})
fig = plt.figure(figsize=(17,10))
ax = []
ax.append(fig.add_subplot(221))
ax.append(fig.add_subplot(222))
ax.append(fig.add_subplot(223))
ax.append(fig.add_subplot(224))
plt.subplots_adjust(hspace=0.7)
years = cwur_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_citations',data=ctry_citations_avg[ctry_citations_avg['year'] == y].head(15), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries rated top in influence¶
Countries like Singapore, Netherlands, Sweden, Israel and Switzerland have instititions that are rated highly in influence. These have only few institutions and they are ranking high due to which they are rated highly. USA and United Kingdom are other countries that have many instututions and are rated high for influence.
In [18]:
# Country wise average influence per institute
ctry_influence_avg = cwur_data.groupby(['year','country'])['influence'].mean()
ctry_influence_avg = pd.DataFrame(ctry_influence_avg)
ctry_influence_avg = ctry_influence_avg.reset_index()
ctry_influence_avg = ctry_influence_avg.sort_values(['year','influence'])
ctry_influence_avg = ctry_influence_avg.rename(columns = {'influence':'avg_influence'})
fig = plt.figure(figsize=(17,10))
ax = []
ax.append(fig.add_subplot(221))
ax.append(fig.add_subplot(222))
ax.append(fig.add_subplot(223))
ax.append(fig.add_subplot(224))
plt.subplots_adjust(hspace=0.7)
years = cwur_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_influence',data=ctry_influence_avg[ctry_influence_avg['year'] == y].head(15), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries rated top in overall score¶
Overall score is a combination of all the features used in ranking i.e. it includes quality of education, quality of faculty, alumni employment, citations, publications and influence. The higher the overall score the better the world ranking of that country.
Countries like Singapore, Netherlands, Russia, Israel and Switzerland have instititions that have high overall score. These have only few institutions and they are ranking high due to which they are rated highly. USA, United Kingdom, Canada and Japan are other countries that have many instututions and have high overall score.
In [19]:
# Country wise average score per institute
ctry_score_avg = cwur_data.groupby(['year','country'])['score'].mean()
ctry_score_avg = pd.DataFrame(ctry_score_avg)
ctry_score_avg = ctry_score_avg.reset_index()
ctry_score_avg = ctry_score_avg.sort_values(['year','score'],ascending=False)
ctry_score_avg = ctry_score_avg.rename(columns = {'score':'avg_score'})
fig = plt.figure(figsize=(17,10))
ax = []
ax.append(fig.add_subplot(221))
ax.append(fig.add_subplot(222))
ax.append(fig.add_subplot(223))
ax.append(fig.add_subplot(224))
plt.subplots_adjust(hspace=0.7)
years = cwur_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_score',data=ctry_score_avg[ctry_score_avg['year'] == y].head(20), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in patents¶
Most of the top institutions in patents are from the USA. Other countries with top ranked institutions in patents are Japan, China and South Korea.
In [20]:
fig = plt.figure(figsize=(18,5))
ax = []
ax.append(fig.add_subplot(141))
ax.append(fig.add_subplot(142))
ax.append(fig.add_subplot(143))
ax.append(fig.add_subplot(144))
#plt.subplots_adjust(hspace=1.0)
i = 0
for y in years:
#sns.set_context('poster',font_scale=0.7)
g = sns.barplot(x='institution',y='patents',data=cwur_data[cwur_data['year']==y].sort_values(['patents'])[['institution','country','patents']].head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in quality of education¶
Most of the top institutions in quality of education are from the USA. Other countries with top ranked institutions in education quality are United Kingdom and France.
In [21]:
fig = plt.figure(figsize=(18,5))
ax = []
ax.append(fig.add_subplot(141))
ax.append(fig.add_subplot(142))
ax.append(fig.add_subplot(143))
ax.append(fig.add_subplot(144))
i = 0
for y in years:
g = sns.barplot(x='institution',y='quality_of_education',
data=cwur_data[cwur_data['year']==y].sort_values(['quality_of_education'])[['institution','country','quality_of_education']].
head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in alumni employment¶
Most of the top institutions in alumni employment are from the USA. Other countries with top ranked institutions in patents are France and Japan
In [22]:
fig = plt.figure(figsize=(18,5))
ax = []
ax.append(fig.add_subplot(141))
ax.append(fig.add_subplot(142))
ax.append(fig.add_subplot(143))
ax.append(fig.add_subplot(144))
i = 0
for y in years:
g = sns.barplot(x='institution',y='alumni_employment',
data=cwur_data[cwur_data['year']==y].sort_values(['alumni_employment'])[['institution','country','alumni_employment']].
head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in quality of faculty¶
Most of the top institutions in quality of faculty are from the USA. France also has few top ranked institutions.
In [23]:
fig = plt.figure(figsize=(18,5))
ax = []
ax.append(fig.add_subplot(141))
ax.append(fig.add_subplot(142))
ax.append(fig.add_subplot(143))
ax.append(fig.add_subplot(144))
i = 0
for y in years:
g = sns.barplot(x='institution',y='quality_of_faculty',
data=cwur_data[cwur_data['year']==y].sort_values(['quality_of_faculty'])[['institution','country','quality_of_faculty']].
head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in publications¶
Most of the top institutions in publications are from the USA. Other countries with top ranked institutions in patents are Canada and United Kingdom.
In [24]:
fig = plt.figure(figsize=(18,5))
ax = []
ax.append(fig.add_subplot(141))
ax.append(fig.add_subplot(142))
ax.append(fig.add_subplot(143))
ax.append(fig.add_subplot(144))
i = 0
for y in years:
g = sns.barplot(x='institution',y='publications',
data=cwur_data[cwur_data['year']==y].
sort_values(['publications'])[['institution','country','publications']].
head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in citations¶
Most of the top institutions in citations are from the USA.
In [25]:
fig = plt.figure(figsize=(18,5))
ax = []
ax.append(fig.add_subplot(141))
ax.append(fig.add_subplot(142))
ax.append(fig.add_subplot(143))
ax.append(fig.add_subplot(144))
i = 0
for y in years:
g = sns.barplot(x='institution',y='citations',
data=cwur_data[cwur_data['year']==y].
sort_values(['citations'])[['institution','country','citations']].
head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in infuence¶
Most of the top institutions in influence are from the USA. United Kingdom also has few top ranked institutions.
In [26]:
fig = plt.figure(figsize=(18,5))
ax = []
ax.append(fig.add_subplot(141))
ax.append(fig.add_subplot(142))
ax.append(fig.add_subplot(143))
ax.append(fig.add_subplot(144))
i = 0
for y in years:
g = sns.barplot(x='institution',y='influence',
data=cwur_data[cwur_data['year']==y].
sort_values(['influence'])[['institution','country','influence']].
head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Multivariate Analysis¶
In [27]:
# Create dataset subset for each year to find correletaion among the numeric variables. National rank column is ignored
cwur_data_2012 = cwur_data[cwur_data['year']==2012][['world_rank','quality_of_education','alumni_employment', 'quality_of_faculty',
'publications', 'influence', 'citations', 'patents', 'score']]
cwur_data_2013 = cwur_data[cwur_data['year']==2013][['world_rank','quality_of_education','alumni_employment', 'quality_of_faculty',
'publications', 'influence', 'citations', 'patents', 'score']]
cwur_data_2014 = cwur_data[cwur_data['year']==2014][['world_rank','quality_of_education','alumni_employment', 'quality_of_faculty',
'publications', 'influence', 'citations', 'patents', 'score']]
cwur_data_2015 = cwur_data[cwur_data['year']==2015][['world_rank','quality_of_education','alumni_employment', 'quality_of_faculty',
'publications', 'influence', 'citations', 'patents', 'score']]
Correlogram plot to see relation between the variables¶
All the years show almost similar correlations. We see that score is strongly negatively related to most of the variables as better (lower value) the ranks in different categories the higher the score is. Quality of faculty, quality of education and world rank are the most determinant factors for score. Quality of faculty and influence are string determinant for world rank.
We also see that publications, citations and influence are strongly positively correlated.
Alumni employment is weakly correlated to most of the variables especially quality of faculty, citations and influence. Similarly quality of education is weakly correlated to citations, publications and patents.
In [28]:
fig = plt.figure(figsize=(15,7))
sns.heatmap(cwur_data_2015.corr(),cmap='coolwarm')
Out[28]:
In [29]:
fig = plt.figure(figsize=(15,7))
sns.heatmap(cwur_data_2014.corr(),cmap='coolwarm')
Out[29]:
In [30]:
fig = plt.figure(figsize=(15,7))
sns.heatmap(cwur_data_2013.corr(),cmap='coolwarm')
Out[30]:
In [31]:
fig = plt.figure(figsize=(15,7))
sns.heatmap(cwur_data_2012.corr(),cmap='coolwarm',annot=True)
Out[31]:
The cluster map shows that quality of education and quality of faculty are in the same group. Citations, publications and influence are close together. Score is away from all other variables
In [32]:
sns.clustermap(cwur_data_2015.corr())
Out[32]:
Heatmap showing countries and instutions ranking¶
In 2012 and 2013 only top 100 institutions were ranked. USA has significant number of top rated institutions in these two years. From 2014 top 1000 institutions were ranked. Even in 2014 and 2015 USA had most number of institutions ranked in top 1000. it is floowed by China, Japan, United Kingdom, Germany and France. This graph shows similar information as shown by graph in "Countries having top universities" section but it has a different view.
In [33]:
ctry_inst = cwur_data.groupby(['year','country'])['institution'].count()
ctry_inst = pd.DataFrame(ctry_inst)
ctry_inst = ctry_inst.reset_index()
pv_ctry_inst = ctry_inst.pivot_table(values='institution',index='country',columns='year')
pv_ctry_inst.fillna(0, inplace=True)
fig = plt.figure(figsize=(15,15))
sns.heatmap(pv_ctry_inst, cmap='Reds',linecolor='white',linewidths=1)
Out[33]:
Wordcloud¶
In [83]:
wc = WordCloud(background_color="white").generate(" ".join([i for i in cwur_data[cwur_data['year'] == 2015]['country']]))
plt.figure( figsize=(10,12))
plt.imshow(wc)
plt.axis("off")
Out[83]:
World Chloropleth Map for top ranked universities¶
In [161]:
init_notebook_mode(connected=False)
In [162]:
ctry_inst_2015 = ctry_inst[ctry_inst['year']==2015]
In [163]:
data = dict(type='choropleth',
locationmode = 'country names',
colorscale = 'Portland',
autocolorscale = False,
reversescale = True,
locations = ctry_inst_2015['country'],
z = ctry_inst_2015['institution'],
text = ctry_inst_2015['country'],
marker = dict(line = dict(color = 'rgb(102,102,102)',width = 2)),
colorbar = {'title':"Institutions count"}
)
In [164]:
layout = dict(
title = '2015 World Class Universities',
geo = dict(
scope='world',
projection=dict( type='robinson' ),
showlakes = True,
lakecolor = 'rgb(74, 129, 179)'
)
)
In [165]:
choromap = go.Figure(data = [data],layout = layout)
iplot(choromap)
Times Ranking Analysis¶
In [34]:
times_data.head()
Out[34]:
In [35]:
## Summary Statistics
times_data.describe()
Out[35]:
In [36]:
# Get dataset dimensions and variable data types
times_data.info()
In [37]:
# Size of dataset
times_data.shape
Out[37]:
In [38]:
times_data.dtypes
Out[38]:
Some Cleaning¶
In [39]:
# International stidents is in percentage. '%' is removed and then the variable is converted into numeric
times_data['international_students'] = times_data['international_students'].map(lambda x: str(x)[:-1])
times_data['international_students'] = times_data['international_students'].fillna(0)
times_data['international_students'].replace(regex=True,inplace=True,to_replace=r'na',value=r'0')
times_data['international_students'] = pd.to_numeric(times_data['international_students'])
In [40]:
times_data['international'].replace(regex=True,inplace=True,to_replace=r'-',value=r'0')
times_data['international'] = pd.to_numeric(times_data['international'])
In [41]:
times_data['income'].replace(regex=True,inplace=True,to_replace=r'-',value=r'0')
times_data['income'] = pd.to_numeric(times_data['income'])
In [42]:
times_data['total_score'].replace(regex=True,inplace=True,to_replace=r'-',value=r'0')
times_data['total_score'] = pd.to_numeric(times_data['total_score'])
In [43]:
times_data['num_students'].replace(regex=True,inplace=True,to_replace=r',',value=r'')
times_data['num_students'] = pd.to_numeric(times_data['num_students'])
times_data['num_students']= times_data['num_students'].fillna(0).astype(int)
In [44]:
temp2 = times_data.world_rank.str.split('-', expand=True)
temp3 = times_data.join(temp2[0])
temp3.drop('world_rank',axis = 1,inplace=True)
times_data = temp3.rename(columns={0:'world_rank'})
times_data['world_rank'].replace(regex=True,inplace=True,to_replace=r'=',value=r'')
times_data['world_rank'] = pd.to_numeric(times_data['world_rank'])
In [45]:
temp2 = times_data.female_male_ratio.str.split(':', expand=True)
temp3 = times_data.join(temp2[0])
temp3 = temp3.join(temp2[1])
temp3.drop('female_male_ratio',axis = 1,inplace=True)
temp3[0].replace(regex=True,inplace=True,to_replace=r'-',value=r'')
temp3[0] = pd.to_numeric(temp3[0])
temp3[1] = pd.to_numeric(temp3[1])
temp3['female_male_ratio'] = temp3[0]/temp3[1]
temp3.drop(0,axis = 1,inplace=True)
temp3.drop(1,axis = 1,inplace=True)
times_data = temp3
Year wise institutions ranked distribution¶
In [46]:
sns.countplot(x='year', data=times_data)
Out[46]:
Countries rated top in teaching¶
Singapore, USA, Brazil and Japan
In [47]:
# Country wise average patents ranking per institute
ctry_teaching_avg = times_data.groupby(['year','country'])['teaching'].mean()
ctry_teaching_avg = pd.DataFrame(ctry_teaching_avg)
ctry_teaching_avg = ctry_teaching_avg.reset_index()
ctry_teaching_avg = ctry_teaching_avg.sort_values(['year','teaching'],ascending=False)
ctry_teaching_avg = ctry_teaching_avg.rename(columns = {'teaching':'avg_teaching'})
fig = plt.figure(figsize=(17,20))
ax = []
ax.append(fig.add_subplot(321))
ax.append(fig.add_subplot(322))
ax.append(fig.add_subplot(323))
ax.append(fig.add_subplot(324))
ax.append(fig.add_subplot(325))
ax.append(fig.add_subplot(326))
plt.subplots_adjust(hspace=1.0)
#plt.subplots_adjust(wspace=0.7)
years = times_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_teaching',data=ctry_teaching_avg[ctry_teaching_avg['year'] == y].head(10), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries rated top in international¶
Singapore, Switzerland, New Zealand and UK
In [48]:
# Country wise average patents ranking per institute
ctry_international_avg = times_data.groupby(['year','country'])['international'].mean()
ctry_international_avg = pd.DataFrame(ctry_international_avg)
ctry_international_avg = ctry_international_avg.reset_index()
ctry_international_avg = ctry_international_avg.sort_values(['year','international'],ascending=False)
ctry_international_avg = ctry_international_avg.rename(columns = {'international':'avg_international'})
fig = plt.figure(figsize=(17,20))
ax = []
ax.append(fig.add_subplot(321))
ax.append(fig.add_subplot(322))
ax.append(fig.add_subplot(323))
ax.append(fig.add_subplot(324))
ax.append(fig.add_subplot(325))
ax.append(fig.add_subplot(326))
plt.subplots_adjust(hspace=1.0)
#plt.subplots_adjust(wspace=0.7)
years = times_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_international',data=ctry_international_avg[ctry_international_avg['year'] == y].head(10), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries rated top in research¶
Singapore, Netherlands, USA, Japan, Canada
In [49]:
# Country wise average patents ranking per institute
ctry_research_avg = times_data.groupby(['year','country'])['research'].mean()
ctry_research_avg = pd.DataFrame(ctry_research_avg)
ctry_research_avg = ctry_research_avg.reset_index()
ctry_research_avg = ctry_research_avg.sort_values(['year','research'],ascending=False)
ctry_research_avg = ctry_research_avg.rename(columns = {'research':'avg_research'})
fig = plt.figure(figsize=(17,20))
ax = []
ax.append(fig.add_subplot(321))
ax.append(fig.add_subplot(322))
ax.append(fig.add_subplot(323))
ax.append(fig.add_subplot(324))
ax.append(fig.add_subplot(325))
ax.append(fig.add_subplot(326))
plt.subplots_adjust(hspace=1.0)
#plt.subplots_adjust(wspace=0.7)
years = times_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_research',data=ctry_research_avg[ctry_research_avg['year'] == y].head(10), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries rated top in citations¶
USA, UK, Switzerland.
In [50]:
# Country wise average patents ranking per institute
ctry_citations_avg = times_data.groupby(['year','country'])['citations'].mean()
ctry_citations_avg = pd.DataFrame(ctry_citations_avg)
ctry_citations_avg = ctry_citations_avg.reset_index()
ctry_citations_avg = ctry_citations_avg.sort_values(['year','citations'],ascending=False)
ctry_citations_avg = ctry_citations_avg.rename(columns = {'citations':'avg_citations'})
fig = plt.figure(figsize=(17,20))
ax = []
ax.append(fig.add_subplot(321))
ax.append(fig.add_subplot(322))
ax.append(fig.add_subplot(323))
ax.append(fig.add_subplot(324))
ax.append(fig.add_subplot(325))
ax.append(fig.add_subplot(326))
plt.subplots_adjust(hspace=1.0)
#plt.subplots_adjust(wspace=0.7)
years = times_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_citations',data=ctry_citations_avg[ctry_citations_avg['year'] == y].head(10), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries rated top in income¶
South Africa, South Korea, Netherlands, China.
In [51]:
# Country wise average patents ranking per institute
ctry_income_avg = times_data.groupby(['year','country'])['income'].mean()
ctry_income_avg = pd.DataFrame(ctry_income_avg)
ctry_income_avg = ctry_income_avg.reset_index()
ctry_income_avg = ctry_income_avg.sort_values(['year','income'],ascending=False)
ctry_income_avg = ctry_income_avg.rename(columns = {'income':'avg_income'})
fig = plt.figure(figsize=(17,20))
ax = []
ax.append(fig.add_subplot(321))
ax.append(fig.add_subplot(322))
ax.append(fig.add_subplot(323))
ax.append(fig.add_subplot(324))
ax.append(fig.add_subplot(325))
ax.append(fig.add_subplot(326))
plt.subplots_adjust(hspace=1.0)
#plt.subplots_adjust(wspace=0.7)
years = times_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_income',data=ctry_income_avg[ctry_income_avg['year'] == y].head(10), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries rated top in total score¶
Singapore, Netherlands, Switzerland, USA< UK
In [52]:
# Country wise average patents ranking per institute
ctry_total_score_avg = times_data.groupby(['year','country'])['total_score'].mean()
ctry_total_score_avg = pd.DataFrame(ctry_total_score_avg)
ctry_total_score_avg = ctry_total_score_avg.reset_index()
ctry_total_score_avg = ctry_total_score_avg.sort_values(['year','total_score'],ascending=False)
ctry_total_score_avg = ctry_total_score_avg.rename(columns = {'total_score':'avg_total_score'})
fig = plt.figure(figsize=(17,20))
ax = []
ax.append(fig.add_subplot(321))
ax.append(fig.add_subplot(322))
ax.append(fig.add_subplot(323))
ax.append(fig.add_subplot(324))
ax.append(fig.add_subplot(325))
ax.append(fig.add_subplot(326))
plt.subplots_adjust(hspace=1.0)
#plt.subplots_adjust(wspace=0.7)
years = times_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_total_score',data=ctry_total_score_avg[ctry_total_score_avg['year'] == y].head(10), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries rated top in number of students¶
Varies with year
In [53]:
# Country wise average patents ranking per institute
ctry_num_students_avg = times_data.groupby(['year','country'])['num_students'].mean()
ctry_num_students_avg = pd.DataFrame(ctry_num_students_avg)
ctry_num_students_avg = ctry_num_students_avg.reset_index()
ctry_num_students_avg = ctry_num_students_avg.sort_values(['year','num_students'],ascending=False)
ctry_num_students_avg = ctry_num_students_avg.rename(columns = {'num_students':'avg_num_students'})
fig = plt.figure(figsize=(17,20))
ax = []
ax.append(fig.add_subplot(321))
ax.append(fig.add_subplot(322))
ax.append(fig.add_subplot(323))
ax.append(fig.add_subplot(324))
ax.append(fig.add_subplot(325))
ax.append(fig.add_subplot(326))
plt.subplots_adjust(hspace=1.0)
#plt.subplots_adjust(wspace=0.7)
years = times_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_num_students',data=ctry_num_students_avg[ctry_num_students_avg['year'] == y].head(10), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries rated top in student staff ratio¶
It varies. Japan is one among them.
In [54]:
# Country wise average patents ranking per institute
ctry_ssr_avg = times_data.groupby(['year','country'])['student_staff_ratio'].mean()
ctry_ssr_avg = pd.DataFrame(ctry_ssr_avg)
ctry_ssr_avg = ctry_ssr_avg.reset_index()
ctry_ssr_avg = ctry_ssr_avg.sort_values(['year','student_staff_ratio'])
ctry_ssr_avg = ctry_ssr_avg.rename(columns = {'student_staff_ratio':'avg_student_staff_ratio'})
fig = plt.figure(figsize=(17,20))
ax = []
ax.append(fig.add_subplot(321))
ax.append(fig.add_subplot(322))
ax.append(fig.add_subplot(323))
ax.append(fig.add_subplot(324))
ax.append(fig.add_subplot(325))
ax.append(fig.add_subplot(326))
plt.subplots_adjust(hspace=1.0)
#plt.subplots_adjust(wspace=0.7)
years = times_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_student_staff_ratio',data=ctry_ssr_avg[ctry_ssr_avg['year'] == y].head(10), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries rated top in international students¶
Singapore, Switzerland, United Kingdom
In [55]:
# Country wise average patents ranking per institute
ctry_is_avg = times_data.groupby(['year','country'])['international_students'].mean()
ctry_is_avg = pd.DataFrame(ctry_is_avg)
ctry_is_avg = ctry_is_avg.reset_index()
ctry_is_avg = ctry_is_avg.sort_values(['year','international_students'],ascending=False)
ctry_is_avg = ctry_is_avg.rename(columns = {'international_students':'avg_international_students'})
fig = plt.figure(figsize=(17,20))
ax = []
ax.append(fig.add_subplot(321))
ax.append(fig.add_subplot(322))
ax.append(fig.add_subplot(323))
ax.append(fig.add_subplot(324))
ax.append(fig.add_subplot(325))
ax.append(fig.add_subplot(326))
plt.subplots_adjust(hspace=1.0)
#plt.subplots_adjust(wspace=0.7)
years = times_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_international_students',data=ctry_is_avg[ctry_is_avg['year'] == y].head(10), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Countries rated top in world rank¶
USA, UK, Singapore, Switzerland
In [56]:
# Country wise average patents ranking per institute
ctry_world_rank_avg = times_data.groupby(['year','country'])['world_rank'].mean()
ctry_world_rank_avg = pd.DataFrame(ctry_world_rank_avg)
ctry_world_rank_avg = ctry_world_rank_avg.reset_index()
ctry_world_rank_avg = ctry_world_rank_avg.sort_values(['year','world_rank'])
ctry_world_rank_avg = ctry_world_rank_avg.rename(columns = {'world_rank':'avg_world_rank'})
fig = plt.figure(figsize=(17,20))
ax = []
ax.append(fig.add_subplot(321))
ax.append(fig.add_subplot(322))
ax.append(fig.add_subplot(323))
ax.append(fig.add_subplot(324))
ax.append(fig.add_subplot(325))
ax.append(fig.add_subplot(326))
plt.subplots_adjust(hspace=1.0)
#plt.subplots_adjust(wspace=0.7)
years = times_data.year.unique()
i = 0
for y in years:
g = sns.barplot(x='country',y='avg_world_rank',data=ctry_world_rank_avg[ctry_world_rank_avg['year'] == y].head(10), ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in teaching¶
Most of the top institutions in teaching are from the USA. United Kingdom also has few top ranked institutions.
In [57]:
fig = plt.figure(figsize=(18,15))
ax = []
ax.append(fig.add_subplot(231))
ax.append(fig.add_subplot(232))
ax.append(fig.add_subplot(233))
ax.append(fig.add_subplot(234))
ax.append(fig.add_subplot(235))
ax.append(fig.add_subplot(236))
plt.subplots_adjust(hspace=1.0)
years = times_data.year.unique()
i = 0
for y in years:
#sns.set_context('poster',font_scale=0.7)
g = sns.barplot(x='university_name',y='teaching',data=times_data[times_data['year']==y].sort_values(['teaching'], ascending=False)[['university_name','country','teaching']].head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in international¶
Top institutions ain international are spread across Switzerlan, United Kingdom, Australia and few other countries.
In [58]:
fig = plt.figure(figsize=(18,15))
ax = []
ax.append(fig.add_subplot(231))
ax.append(fig.add_subplot(232))
ax.append(fig.add_subplot(233))
ax.append(fig.add_subplot(234))
ax.append(fig.add_subplot(235))
ax.append(fig.add_subplot(236))
plt.subplots_adjust(hspace=1.0)
years = times_data.year.unique()
i = 0
for y in years:
#sns.set_context('poster',font_scale=0.7)
g = sns.barplot(x='university_name',y='international',data=times_data[times_data['year']==y].sort_values(['international'], ascending=False)[['university_name','country','international']].head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in research¶
USA and UK institutes top in research.
In [59]:
fig = plt.figure(figsize=(18,15))
ax = []
ax.append(fig.add_subplot(231))
ax.append(fig.add_subplot(232))
ax.append(fig.add_subplot(233))
ax.append(fig.add_subplot(234))
ax.append(fig.add_subplot(235))
ax.append(fig.add_subplot(236))
plt.subplots_adjust(hspace=1.0)
years = times_data.year.unique()
i = 0
for y in years:
#sns.set_context('poster',font_scale=0.7)
g = sns.barplot(x='university_name',y='research',data=times_data[times_data['year']==y].sort_values(['research'], ascending=False)[['university_name','country','research']].head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in citations¶
Most of the top institutions in citations are from the USA.
In [60]:
fig = plt.figure(figsize=(18,15))
ax = []
ax.append(fig.add_subplot(231))
ax.append(fig.add_subplot(232))
ax.append(fig.add_subplot(233))
ax.append(fig.add_subplot(234))
ax.append(fig.add_subplot(235))
ax.append(fig.add_subplot(236))
plt.subplots_adjust(hspace=1.0)
years = times_data.year.unique()
i = 0
for y in years:
#sns.set_context('poster',font_scale=0.7)
g = sns.barplot(x='university_name',y='citations',data=times_data[times_data['year']==y].sort_values(['citations'], ascending=False)[['university_name','country','citations']].head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in income¶
USA, Netherlands and South Korea have top 10 universities in income.
In [61]:
fig = plt.figure(figsize=(18,15))
ax = []
ax.append(fig.add_subplot(231))
ax.append(fig.add_subplot(232))
ax.append(fig.add_subplot(233))
ax.append(fig.add_subplot(234))
ax.append(fig.add_subplot(235))
ax.append(fig.add_subplot(236))
plt.subplots_adjust(hspace=1.0)
years = times_data.year.unique()
i = 0
for y in years:
#sns.set_context('poster',font_scale=0.7)
g = sns.barplot(x='university_name',y='income',data=times_data[times_data['year']==y].sort_values(['income'], ascending=False)[['university_name','country','income']].head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in total score¶
Top 10 universities in total score are from USA and UK.
In [62]:
fig = plt.figure(figsize=(18,15))
ax = []
ax.append(fig.add_subplot(231))
ax.append(fig.add_subplot(232))
ax.append(fig.add_subplot(233))
ax.append(fig.add_subplot(234))
ax.append(fig.add_subplot(235))
ax.append(fig.add_subplot(236))
plt.subplots_adjust(hspace=1.0)
years = times_data.year.unique()
i = 0
for y in years:
#sns.set_context('poster',font_scale=0.7)
g = sns.barplot(x='university_name',y='total_score',data=times_data[times_data['year']==y].sort_values(['total_score'], ascending=False)[['university_name','country','total_score']].head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in number of students¶
USA and Italy have few institutes that are in top 10 in terms of total students intake.
In [63]:
fig = plt.figure(figsize=(18,15))
ax = []
ax.append(fig.add_subplot(231))
ax.append(fig.add_subplot(232))
ax.append(fig.add_subplot(233))
ax.append(fig.add_subplot(234))
ax.append(fig.add_subplot(235))
ax.append(fig.add_subplot(236))
plt.subplots_adjust(hspace=1.0)
years = times_data.year.unique()
i = 0
for y in years:
#sns.set_context('poster',font_scale=0.7)
g = sns.barplot(x='university_name',y='num_students',data=times_data[times_data['year']==y].sort_values(['num_students'], ascending=False)[['university_name','country','num_students']].head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions in student staff ratio¶
USA, Japan, Sweden and Denmark have few institutes with verly low student staff ration
In [64]:
fig = plt.figure(figsize=(18,15))
ax = []
ax.append(fig.add_subplot(231))
ax.append(fig.add_subplot(232))
ax.append(fig.add_subplot(233))
ax.append(fig.add_subplot(234))
ax.append(fig.add_subplot(235))
ax.append(fig.add_subplot(236))
plt.subplots_adjust(hspace=1.0)
years = times_data.year.unique()
i = 0
for y in years:
#sns.set_context('poster',font_scale=0.7)
g = sns.barplot(x='university_name',y='student_staff_ratio',data=times_data[times_data['year']==y].sort_values(['student_staff_ratio'])[['university_name','country','student_staff_ratio']].head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
Top 10 institutions having high percentage of international students¶
UK and Netherlands have high percentage of international students.
In [65]:
fig = plt.figure(figsize=(18,15))
ax = []
ax.append(fig.add_subplot(231))
ax.append(fig.add_subplot(232))
ax.append(fig.add_subplot(233))
ax.append(fig.add_subplot(234))
ax.append(fig.add_subplot(235))
ax.append(fig.add_subplot(236))
plt.subplots_adjust(hspace=1.0)
years = times_data.year.unique()
i = 0
for y in years:
#sns.set_context('poster',font_scale=0.7)
g = sns.barplot(x='university_name',y='international_students',data=times_data[times_data['year']==y].sort_values(['international_students'], ascending=False)[['university_name','country','international_students']].head(10), hue='country', ax=ax[i])
g.axes.set_title(y, fontsize=20,color="r",alpha=0.5)
for item in g.get_xticklabels():
item.set_rotation(90)
i += 1
In [66]:
times_data_2011 = times_data[times_data['year']==2011][['teaching', 'international','research', 'citations', 'income',
'total_score', 'num_students','student_staff_ratio',
'international_students', 'world_rank', 'female_male_ratio']]
times_data_2012 = times_data[times_data['year']==2012][['teaching', 'international','research', 'citations', 'income',
'total_score', 'num_students','student_staff_ratio',
'international_students', 'world_rank', 'female_male_ratio']]
times_data_2013 = times_data[times_data['year']==2013][['teaching', 'international','research', 'citations', 'income',
'total_score', 'num_students','student_staff_ratio',
'international_students', 'world_rank', 'female_male_ratio']]
times_data_2014 = times_data[times_data['year']==2014][['teaching', 'international','research', 'citations', 'income',
'total_score', 'num_students','student_staff_ratio',
'international_students', 'world_rank', 'female_male_ratio']]
times_data_2015 = times_data[times_data['year']==2015][['teaching', 'international','research', 'citations', 'income',
'total_score', 'num_students','student_staff_ratio',
'international_students', 'world_rank', 'female_male_ratio']]
times_data_2016 = times_data[times_data['year']==2016][['teaching', 'international','research', 'citations', 'income',
'total_score', 'num_students','student_staff_ratio',
'international_students', 'world_rank', 'female_male_ratio']]
Correlogram plot to see relation between the variables¶
Across all the years we see that
- World rank is strongly negatively correlated with many variables. Citations, research, teaching and overall score drive the universities' world ranking.
- Number of students and student staff ratio are weakly associated with most of the variables. They are moderately positively correlated only to each other.
- Research is strongly positively correlated to teaching and total score. Teaching and research highly drives total score.
- International students and international are strongly positively correlated
- Female male ratio is also weakly related to most of the variables. Female male ratio, Number of students and student staff ratio do not influence total score and world ranking
In [67]:
fig = plt.figure(figsize=(15,7))
sns.heatmap(times_data_2011.corr(),cmap='coolwarm')
Out[67]:
In [68]:
fig = plt.figure(figsize=(15,7))
sns.heatmap(times_data_2012.corr(),cmap='coolwarm')
Out[68]:
In [69]:
fig = plt.figure(figsize=(15,7))
sns.heatmap(times_data_2013.corr(),cmap='coolwarm')
Out[69]:
In [70]:
fig = plt.figure(figsize=(15,7))
sns.heatmap(times_data_2014.corr(),cmap='coolwarm')
Out[70]:
In [71]:
fig = plt.figure(figsize=(15,7))
sns.heatmap(times_data_2015.corr(),cmap='coolwarm')
Out[71]:
In [72]:
fig = plt.figure(figsize=(15,7))
sns.heatmap(times_data_2016.corr(),cmap='coolwarm',annot=True)
Out[72]:
The cluster map shows that -
- student staff ratio and number of students are in the same group
- international students and international are in the same group
- research and teaching are in the same group
- world rank is a combination of all the factors
In [73]:
sns.clustermap(times_data_2016.corr())
Out[73]:
Heatmap showing countries and universities ranking¶
Across all the years we see USA tops the list with maximum number of top ranked universities. This is followed by United Kingdom. Other countries having large number of top ranked universities are Germany, Japan, Italy, Australia, France, Canada and China.
In [74]:
ctry_univ = times_data.groupby(['year','country'])['university_name'].count()
ctry_univ = pd.DataFrame(ctry_univ)
ctry_univ = ctry_univ.reset_index()
pv_ctry_univ = ctry_univ.pivot_table(values='university_name',index='country',columns='year')
pv_ctry_univ.fillna(0, inplace=True)
fig = plt.figure(figsize=(15,15))
sns.heatmap(pv_ctry_univ, cmap='Reds',linecolor='white',linewidths=1)
Out[74]:
